Investigating the Farallon Slab with Probabilistic Traveltime Tomography

Abstract:

Subduction of the Farallon Plate beneath North America played a key role in its tectonic development. Seismic constraints on the subducted remnants of the Farallon slab provide evidence needed to better understand the polarity and timing of subduction, the structure of the plate, and its relation to tectonic events like the uplift of the Rocky Mountains. Over the course of its deployment, the USArray Transportable Array (TA) has offered ideal data coverage for investigating the Farallon and related slabs in the upper mantle using seismic tomography and converted wave imaging. With its arrival in the east, data from the TA provides the crossing paths necessary to image the upper reaches of the oldest parts of the plate at mid-mantle depths.

We perform a global tomographic inversion using the latest P-wave traveltime picks from TA combined with global catalogue data. While the new velocity model resolves upper mantle slab structure at unprecedented detail in the east, a quantitative grasp of model uncertainty is needed to reliably relate velocity variations to the thermal and mechanical properties of the slabs. In order to quantify the uncertainty of our tomographic model, we employ Transdimensional Hierarchical Bayesian (THB) inversion. THB tomography uses Markov chain Monte Carlo to create an ensemble of velocity models that can be analyzed to statistically infer the best-fit velocities, their uncertainties, and tradeoffs.

We present and discuss various representations of uncertainty quantified by THB tomography—error bars, model covariance, multimodal distributions of velocity values—and demonstrate its importance for furthering our understanding of the slab fragments beneath North America. We illustrate how we are able to distinguish between spurious slab fragments from those required by the data. By examining bimodal velocity distributions, we put error bars on the spatial extent of the slabs that can then be analyzed using thermal diffusion modeling. By mapping out regions of high model covariance, we quantify smearing in the tomographic model, which allows us, for example, to assess the extent to which the deep slab fragments near the coasts are artifacts of data coverage. Preliminary results cast doubt on the interpretability of fast anomalies in the upper mantle off the west coast of the continent.